• Modeling Phenotypes for comparative biology is our first focus. We design methods to automate the processes of modeling, integration, organization, mining and visualization of non-molecular phenotypic data and knowledge. We also adopt a multidisciplinary approach (informatics, genomics and natural language processing) to explore the value of semantic, probabilistic and terminological technologies in phenotypic data and to construct contextual networks of phenotypes. Main research thrust: Phenotype Organizer System [POS, Phenoslim-SNOMED, OMIM-SNOMED, QMR-OMIM, GO-UMLS, BiomedLEE online].
• Comparative Phenomics, understanding phenotypic-genotypic interactions , is the second major focus of our research group. We conduct detailed analyses of integrated clinical and functional genomic datasets (clinical genomics), which are closely coupled with our work on mining networks of non-molecular phenotypes and on integration of databases. Examplar projects: Clinigenes/Genestrace, Phenogenes Viewer.
• Integrating Biomedical Datasets across heterogeneous and semi-structured databases is our third complementary focus. To increase the accuracy and accelerate the development of integrated datasets, we propose original model theoretic approaches combined with multiple established methods (structured ontologies, computational terminologies, lexical and semantic technologies, predicate logic calculus, natural language understanding, etc.). Examplar project: Phenotype Organizer System.

Summary Statement. The Lussier Research Group conducts research in the emerging field of systems biology, using computation to model phenotypes, integrate genomic with phenotypic datasets, and analyze phenomes in order to accurately individualize the understanding, the prediction, and the treatment of diseases.
Our research group recently published breakthrough studies in phenomics: computational methods that automatically predict and organize related phenotypes between species and across heterogeneous phenotypic datasets. While reverse genetic studies can infer, in high throughput, relationships between phenotypes and pathogenesis through genetic homology, the "phenotype gap" between clinical and biological databases precludes the converse: high throughput forward genomic studies relating genes to non molecular pathogenic processes. Indeed, specifying an observed phenotype and comparing it to related ones from other organisms remains challenging and is currently conducted manually with curators, a rate limiting, time consuming and expensive endeavor. The proposed high throughput methods efficiently bridge the "phenotype gap", an essential pathway for comparative studies of phenotypes in systems biology, which is likely to impact our understanding of the phenome and consequently of medicine.
The following two publications describe components of the method:
• Cantor MN, *Lussier YA**. “Mining OMIM for Insight in Complex Diseases”. Medinfo 2004, Selected Paper for the International Journal of Medical Informatics (In press, selected for one of the best symposium communication and publication in the Interantional Journal of medical Informatics)
• Lussier YA*, Li J. Terminological Mapping For High Throughput Comparative Biology of Phenotypes. Pacific Symposium on Biocomputing, 2004:202-13.

Specific research projects include:
A) Clinical Genomics Technology
The Human Genome has set the pace for post-genomic discovery research. While post-genomic fields focused at the molecular level are intensively pursued, little effort is being deployed in the later stages of molecular medicine discovery research, such as Clinical (functional) Genomics. The following pioneering studies aim at demonstrating the relevance and significance of integrating mainstream clinical informatics science to current bioinformatics genomic discovery science:

1. Phenotype Organizer System (POS):
   The long-term goal of this NIH-funded project is to build innovative informatics tools, capable of automatically querying, organizing and visualizing phenotypic data (traits, syndromes, etc.) across Phenotype databases, to facilitate phenotypic research that aims to unlock the gene-disease-relationships. This program proposes to adopt a multidisciplinary approach (informatics, genomics and biomedical research) to explore the value of semantic, probabilistic and terminological technologies in phenotypic data and knowledge processing. The proposed research may provide a unique approach to accelerate biomedical research by improving access to phenotypic data and knowledge processing – the Semantic Phenome (phenotypic-genomic relations. We have conducted several proof-of-concept studies (e.g., QMR-OMIM, GenesTrace).
2. Modeling of Emerging Infectious Diseases
   In collaboration with Ian Lipkin, Mark Gerstein, Jeffery Skolnick and Andrea Califano, we are conceptualizing the framework and developing the "PathoGene" software platform for the molecular and clinical modeling of EID. FPDS, The Foundational Pathogen Database System, is a system that will provide improved interoperable access to phenotypic and molecular data about host-pathogen interactions across otherwise isolated database silos.
   * ICTVdb, hosted in our Informatics Core of the NBC, and SNOMED are two of the databases we are interoperating,
3. Molecular Medicine Matrix M³
   The Molecular Medicine Matrix is a project that leverages mediated schemas, language understanding and ontology to enable the creative interoperation of otherwise heterogeneous biological and clinical databases. We have created dynamic maps between a large set of terminologies as GO, OMIM, SNOMED CT, UMLS, PhenoSlim, NCBI and MP (1,3). We are currently adapting M3 in a Phenotype Organizer System (POS) to accelerate comparative biology of phenotypes. Ancestry Analyzer
4. PhenoGenes
   In collaboration with Carol Friedman, we are developing a representational model that depicts genotypic and phenotypic relations found in the literature and also in clinical reports. The knowledge bases of PhenoGenes will be integrated in the Clinigene discovery platform.

B) Individualized medicine Technologies
Lately, the development of clinical practice guidelines (CGPs) and decision support systems (DSS) have received increased emphasis. However, despite this focus on development, less attention has been paid to their integration and evaluation in a genuine clinical practice. The overall goal of this proposal is to develop a modular, portable and multi-institutional DSS supporting CPGs that will improve healthcare. The unique architecture of the Vigilens DSS [PDF] will provides for server-based/tele- event, guideline and outbreak monitoring. Several specific projects stem out of the Vigilens Health Monitor endeavor and are aimed at evaluating the appropriateness and the potential misuse of practice guidelines:

1. Personalized Notification Subsystem
   This funded program is aimed at increasing, evaluating and quantifying the clinical applicability, complexity and flexibility of guidelines including institution's policies and users preferences. We are currently collaborating with IBM Research (Watson Lab) on pervasive notification and Biodefense / Homeland Security applications
2. Rx/Dx
   This project is directed at improving the quality of medication prescribing by personalizing a CGP for the clinical context of an individual patient, taking into account a thorough understanding of their narrative records with language understanding tools and exceptions to the guideline.

Research Projects as co-investigator:

• PhenoGenes Friedman/PI) that uses language understanding to produce advanced knowledge bases in molecular medicine using journal papers,(e.g. Phenogenes Viewer [ //www.dbmi.columbia.edu/~yit7001/files/BM_viewerW.exe ] )
• VigiLens (Shortliffe/PI, Lussier/Co-PI, Mendonça, Johnson) a versatile clinical decision support system, based on a unique architecture, providing event, guideline and outbreak monitoring, VigiLens secure portal
• MI-HEART clinical trial, (Cimino/PI, Lussier, Kukafka, Patel) a computer system that uses patient-specific information from an electronic medical record to produce personalized educational material.
• "Unlocking of Data" with MedLEE (Friedman/PI, Lussier).
  

  [Directions to Lussier's Lab]